Eiji Kudo

Designing Highly Available Infrastructure at Scale

Led the design of EKS and ECS runtime platforms serving several million monthly users and handling thousands of requests per second. Used HPA, Cluster Autoscaler, load testing, and database connection budgets to define safe scaling limits and pre-scaling plans, maintaining availability during traffic spikes of roughly 1,000 times normal levels.

Led stakeholder alignment, budget approval, and the zero-downtime migration of multiple production services from EKS to ECS as the business phase and operating model evolved.

Reduced infrastructure operations and management labour costs by roughly 80%, and combined AWS, GCP, and Datadog cloud and monitoring spend by roughly 50% through continuous usage analysis and resource right-sizing.

Scaling Delivery with Type Safety, Automated QA & CI/CD

Revamped the development and release platform, increasing deployments per engineer by roughly four times and reducing deployment lead time by roughly 30%.

Led product-wide type-safety improvements, including automatically generating typed frontend HTTP clients from backend type definitions.

Established automated testing and AI-driven E2E testing, along with AI usage guidelines, implementation standards, and an agreed review workflow, reducing QA and review effort while improving maintainability.

Designing, Building & Scaling AI Coding Infrastructure Across the Team

Codified project-specific coding conventions, architectural design intent, and prohibited patterns in AGENTS.md. Documented both abstract principles and concrete rules, building a foundation that enables AI to make appropriate decisions even in unfamiliar cases.

Designed and implemented 25+ Custom Skills to automate PR review, planning, and quality check phases. Built a skill suite covering the entire development cycle — including parallel critical PR analysis by specialised reviewers, exhaustive similar-issue detection, QA checklist generation, and sequential lint → type check → test execution with auto-fix.

Built infrastructure to auto-deploy isolated preview environments per PR for rapid verification of AI-generated PRs. Reviewers can start validation by simply accessing a URL, enabling simultaneous verification of multiple AI-generated PRs.

Designing & Building Backend Systems with Managed Services

Built a shared asynchronous processing platform spanning infrastructure and application code, combining queues and Redis according to workload requirements to move high-load and long-running work out of synchronous request paths.

Standardised retries and error handling so workloads could recover automatically from transient failures, then enabled the wider team to apply the same patterns consistently.